time bug

[Stephen Weeks]

> In trying to track down the time bug, I've been running repeated
> self-compiles overnight, bouncing back and forth between stracing and not,
> and saving any logs with errors.  I got another one without stracing:
> 
> mlton: Time:traceBatch: t' - t::-: t1 = 590.720 t2 = 591.710
> 
> Something that might be interesting, both times t1 and t2 differ by 0.99.

Interesting.  Henry, maybe you could take one more look through the kernel with
an eye to this?  Matthew, to get more data but without stracing, you could save
the last few calls to getrusage from the ML side, and only print them if there
is an error.

> Also, I discovered a minor bug in the code I sent out a few days ago.  The
> handle Time wasn't handling the Time exception, because that exception is
> not at the top level.  In fact, it's not even exported by
> /src/lib/mlton/basic/time.sig.  I added the exception to the library TIME
> sig and the exception above really is the Time exception.

Blech.  I added the fix to my stuff.